Discussion Triple architecture CPU or dissimilar dual socket motherboard

Jul 27, 2020
16,288
10,328
106
Intel P-core - Really good single threaded performance

Intel E-core - Acceptable and scalable multicore performance per area

AMD Zen 3+ - Unmatched multicore performance per watt

We have a problem. We want all three types of cores in our PC. We don't want to buy two different PCs and juggle our tasks between them based on which CPU core is best for our use case.

Here's what needs to happen. Intel needs to reach out to AMD and propose a combined CPU with all three cores. AMD shouldn't mind licensing their core to Intel since this arrangement would free them from the burden of catering to desktop users and they can simply focus on their server clients. AMD gets royalties for their CPU cores used in this combined CPU assembled by Intel.

If this is not possible due to disagreement between Intel and AMD, then someone like ASUS needs to build a dual socket mobo incorporating both Intel and AMD chipsets. A KVM solution on the mobo would allow switching between the two systems. It can be a full ATX mobo. For PC users like us, we get the best of both worlds in a relatively compact PC and enjoy less hassle plus time saved from not having to daydream about the other side of the fence. I know this seems like a huge and almost insurmountable engineering problem but I have confidence that the best of the best can make this happen. They just need to believe that something like this would be well received. Please show your support for one or both of these crazy ideas and help make this miracle happen.
 

amd6502

Senior member
Apr 21, 2017
971
360
136
free them from the burden of catering to desktop users and they can simply focus on their server clients.

It would do the opposite, since performance desktop has always piggy backed from the same core design and almost always even the same dies.

Plus, where would that leave desktop users who prefer to go with AMD.

My thought is that the ability to squeeze greater and greater number of cores on the dies makes asymmetric multithreading a good option to have besides SMT2. One strong thread and two or three background threads with optionally selectable non-speculative execution.

Dual socket board that takes both AMD and Intel is a crazy but intriguing idea!! It'd appeal to quite a few enthusiasts, but it sounds like a wild and crazy beast. And makes you wonder of the things that might go wrong.
 

amd6502

Senior member
Apr 21, 2017
971
360
136
Us enthusiasts are no strangers to that, are we? :D

The biggest hurdle might be the limited target market. If the design and support (including software) is complex on top that makes it less likely to happen.

To simplify things, one might just fit two independent systems onto a single motherboard, either matx or atx sized.

They would have totally independent IO, including memory, sata, audio/video out and usb.

matx sized would have a "big" system with almost a full set of usual IO, (two ram slots, 1 full pcie, 1 small pcie, m.2, 4x sata, 8x usb, audio i/o, ethernet). little system, powered by atom, would be minimal (single ram slot, 4x usb, 2x sata, hdmi, audio i/o, ethernet). Users would optionally get KVM switch.

for atx sized you could put two sets of what you normally find in an matx board.

In this case instead of having a soldered on atom soc as in matx example above you'd have the big,little Intel lga socket running one half, and the AM4 socket on the other half (four memory slots total, 2 + 2)
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,554
14,510
136
Well, now that I own an Alderlake, I still can not appreciate the big.little. I think AMD's approach of all cores the same and efficient is the way to go.

Also, under windows 11, it seems very unresponsive for some reason. Way worse than linux on the same box. I mean when I move the mouse to wake it, and its not even running anything, it takes seconds, possibly at times up to a minute for the screen to come back. Whats that about ?
 

amd6502

Senior member
Apr 21, 2017
971
360
136
I think linux has added big.little topology to its kernel since a year or more maybe. It sounds like windows has catching up to do.

Something is really wrong with a minute delay from sleep. Try disabling hybrid sleep where it backs RAM up onto the hard drive. If you have a lot of GB of RAM that might cause significant delay. Other possibility is that the motherboards are just not ready for it and need a lot of work in BIOS updates.

As for when the applications start really raking advantage of the big.little, that might take years. It's going to be a while still till something like a 6c or 8c SMT2 box gets outdone by uplift from a bunch of little cores used effectively.

As for dual socket boards, I really think putting two systems on one board is the way to go, and it would have quite a bit of appeal to users who mostly work in linux and occasionally use windows, or vice versa. Virtual box is an alternative but it is not as good, and seamless mode is very hard to set up.
 
  • Like
Reactions: igor_kavinski

Thibsie

Senior member
Apr 25, 2017
749
801
136
Well, if getting RAM back from hard drive with 8GB was OK, I expect 64GB from nvme drive to RAM should be a piece of cake.
If not, there's a problem indeed.
 

mikeymikec

Lifer
May 19, 2011
17,705
9,566
136
Something is really wrong with a minute delay from sleep. Try disabling hybrid sleep where it backs RAM up onto the hard drive. If you have a lot of GB of RAM that might cause significant delay. Other possibility is that the motherboards are just not ready for it and need a lot of work in BIOS updates.

All that hybrid sleep does is take a backup of the contents of RAM so that if the computer were for example to lose power while in sleep mode, the system can be resumed from the hibernation file. If the computer slept and resumes normally then the hibernation file's contents are ignored.

The only delay that occurs with hybrid sleep is the process of going into sleep mode, not out of it (unless of course power is lost during sleep mode).

I don't know how >Win7 reacts when coming out of hibernation, but Win7 and Vista IIRC both put 'resuming' on the screen with a completely different aura-themed splash screen so it was very obvious when it happened.

----

If I were @Markfw , I'd start by checking the logs at the time of system resume to see whether it complains about anything in particular. My bet is on a display driver reset.

If the event logs reveal nothing of interest, then there's a power system troubleshooting tool in modern versions of Windows (I'd google for Windows sleep troubleshooting to find out the ins and outs of it), IIRC some command switches of powercfg are needed to have Windows analyse the sleep/resume process and give its findings.
 
Last edited:
Jul 27, 2020
16,288
10,328
106
DFI made dissimilar-socket boards, they made an X48 board with an LGA 775 socket on it, that was also combo'ed with a low-power Atom-based system, on the same motherboard. I had a chance to buy one, once, at the MIT Flea.
Wow. What's the model number? Any review link? How was it in practice? Did both run at the same time or was it only boot selection type of thing? Very intriguing!
 
  • Like
Reactions: amd6502
Jul 27, 2020
16,288
10,328
106
I believe that it did. The video out and mouse/keyboard were interesting, I think that it had two sets, but also a sort of pass-through (?).

One was for always-on low-power downloads and stuff, the other was or high-power gaming.
What did you do with it? Is it no longer in your possession now? It's a piece of computing history!
 

nicalandia

Diamond Member
Jan 10, 2019
3,330
5,281
136
So if you are right, what is the area and performance of 16 ecores, compared to a 5950x ? The 5950 is like 2 70 or 80mm dies ? There is no way they can touch a 5950x in performance.


Okay here is is:

Area of 16C/32T of Zen3(5950X) is about 160 mm^2 with L2$ and L3$(not counting the I/O Die) They get about 27,000 points in CB R23 MT at stock settings?

Area of 48C/48T Gracemont cores is about 158 mm^2 with L2$ and L3$, They would get about 45,000 points in CB R23 MT at stock settings.

1647021083628.png

Based on Locuza Die Size and Annotations 16C/32T Gracemont are 52.68 so 48C/48T are about 158 mm^2

1647022193855.png

Updated with correct numbers.....
 
Last edited:
  • Like
Reactions: igor_kavinski

VirtualLarry

No Lifer
Aug 25, 2001
56,339
10,044
126
What's the model number?
Edit: here's a thread about it on another forum.

 
Last edited:
Jul 27, 2020
16,288
10,328
106
Area of 32C/32T Gracemont cores is about 158 mm^2 with L2$ and L3$, They would get about 31,000 points in CB R23 MT at stock settings.
We won't really know coz Intel won't make something like that. I think the E-core is missing something essential which prevents it from booting an OS on its own.
 

nicalandia

Diamond Member
Jan 10, 2019
3,330
5,281
136
We won't really know coz Intel won't make something like that. I think the E-core is missing something essential which prevents it from booting an OS on its own.
Just updated the info with the correct data. Actually at 158 mm^2 we get 48C/48T for a total of 45,000 points in CB R23.. We will see that on the Sierra Forest in a few years